My research focuses on developing image and data analysis software for high throughput live single-cell imaging studies. With this software I am able to extract and model time-series data. This allows us to understand how dynamical signalling networks regulate cell fate. I work with the Dynamical Cell Systems team at the Institute of Cancer Research, supervised by Dr. Chris Bakal, as well as the department of Computational Systems Medicine at Imperial College London, supervised by Prof. Robert Glen
Life sciences are growing rapidly and the desire to be part of this motivated my biochemistry degree from Imperial College. I’ve always enjoyed maths and programming; this has allowed me to take a highly quantitative approach to studying biological systems and is why my PhD is now so cross disciplinary.
Outside of work, I love windsurfing, cycling, skiing and general adventuring. I’ve worked several seasons abroad as a water sports instructor, and previously run the Imperial College windsurfing club.
The majority of our understanding of current biology has been the result of experimenters developing and testing individual hypotheses based on expert knowledge of the problem, or more general question, being addressed. More recently, top down systems biology approaches have sought to understand how molecular interaction networks behave at a ‘system-wide’ level, by simultaneously measuring vast numbers of components. However, in my opinion such methods have largely failed to provide the more complete level of understanding they were suggested to be capable of. Additionally, I think key statistical challenges in performing huge numbers of hypothesis tests simultaneously have also underpinned the current ’reproducibility crisis’ in life sciences.
To address these key challenges I believe we need a return to science driven by individual hypothesis. Though, with the increasingly complex machine learning algorithms now available, we are edging towards a position in which we may be able to automate the process of knowledge acquisition and hypothesis formulation. As such, an overarching goal of my research is the development of pipelines and workflows which will ultimately allow us to take a step back and allow algorithms to produce reproducible results, identify novel behaviour, and ultimately generate better understanding of how biological systems function.
Live single-cell imaging studies allow us to see behaviour and dynamics that would otherwise be masked by methods that average over populations, such as western blotting or mass spectrometry. Two key examples in the image above highlight how a population averaging measure would miss bifurcation of activity into two groups, or sudden switch like dynamics with stochastic timing. However, slowing the uptake of live-single cell methodologies are difficulties in reliably tracking large numbers of often highly motile cells.
The current focus of my PhD is the development of methods capable of reliably tracking single-cells over long periods of time, where imaging is infrequent so as to minimise problems of phytotoxicity, and maximise the number of fields of view that can be imaged in a live cell screen. I have thus developed an integrated tracking package which uses global optimisation methods and a dynamic programming approach to select the best set of tracks from the entire imaging sequence.
By then extracting features from a fluorescently labelled protein (PCNA) in the nucleus I am able to delineate precisely when the cell enters different phases of the cell cycle, and relate this to other fluorescently labelled proteins in the cell. This is allowing us to study the how DNA damage is regulating cell cycle progression in live single-cells.
Also, check out my recent review of the subject Accelerating live single cell signalling studies , published in Trends in Biotechnology.
High-content imaging screens are often used to identify biologically active compounds. To date the majority of morphological screens have been performed on fixed cells, plated on hard substrates. However, many proteins may only be active when cells are in soft environments as often found in the body. For example, melanoma cells migrating in soft tissue demonstrate amoeboid migration; this involving different pathways to the mesenchymal migration mode observed on hard substrates. As a proof of principle, I performed an unsupervised analysis of how single melanoma cells explore different morphologies in 3D collagen environments. Here finding that they transition between mesenchymal and amoeboid morphologies, through different routes. Importantly, the live analysis demonstrated that by taking into account the dynamics of how these cells change shape, the ability to discriminate between different test conditions was greatly improved. Meaning that, short time series measurements may be of value when performing high content screens.
Cooper, Sam, et al. "Apolar and polar transitions drive the conversion between amoeboid and mesenchymal shapes in melanoma cells." Molecular biology of the cell 26.22 (2015): 4163-4170.
Whilst the central focus of my PhD is to develop improved workflows for live single-cell image analysis, within the Bakal lab I am also extensively involved with analysing data from fixed single-cell morphological profiling screens. Here, I have used multiple data analysis methods including principal and canonical component analysis to manage screening batch effects and extract meaningful biological conclusions form large imaging datasets. In a standard workflow as described in the image above, multiple conditions are tested and imaged in high throughput. Subsequently using commercial, or custom scripts which I have developed , cells and their nuclei are identified and segmented. A large number of features are then extracted and these are aggregated over wells, using averaging or clustering methods. Analysis of how different perturbations effect morphology and relate to one another can then identify novel effectors or inhibitors, as well as giving network-wide insights into the system being studied.
As well as research work I also teach a Python course made up of 10 lessons, introducing the language and some core data analysis libraries. The lessons plus associated problems are below, and require the JuPyter notebook to run:
2014 – 2017:   PhD Computational Molecular Biology,   Imperial College London,   Institute for Cancer Research   (3rd year)
Supervisors:   Dr. Chris Bakal ,   Dynamical cell systems team,   Institute of cancer research;   Prof. Robert Glen,   Section of computational systems medicine,   Imperial College London.
Title:   Relating Chemotype to Phenotype;   Data driven approaches to molecular biology
2011-2014:  Imperial College London,   Bsc Biochemistry:   1st Class Honours
2009-2011:   A-levels:   A* Maths,   A* Physics;   A Chemistry;   B Further Maths.   AS-level:   B Biology.   Awarded prize for best in physical sciences
2004-2009,   GCSE:   17 A*-C’s,   Awarded Xyratex prize for best in maths and science.
As well as academic work, I love the more adventurous sports. I’m a qualified sailing and windsurfing instructor. With these qualifications during the summer I’ve worked abroad with three fantastic companies:
I also love skiing in the winter, and cycling at home in the UK.